simulannealbnd gets stuck in an objective function value but is still running

Actually I'm doing SMM (Simulated method of momets), the objective funciton value has been unchanged for many iterations. Is it possible that such occasion to happen? What should I do next?

Answer 1

[Raghav]

Hi Bairu,

Yes, it is possible for the objective function value to remain unchanged for several iterations in the Simulated Method of Moments (SMM) algorithm. This situation can occur due to various reasons, such as the algorithm getting stuck in a local minimum or the optimization process reaching a plateau.

If you encounter this issue, here are a few steps you can take:

1. Check your implementation: Double-check your code and make sure there are no errors or bugs that might be causing the stagnation. Verify that you are correctly calculating the moments and updating the parameters.
2. Adjust the optimization algorithm: Try using a different optimization algorithm or adjusting the parameters of the current algorithm. For example, you can experiment with different learning rates, step sizes, or convergence criteria. Sometimes, a small tweak can help the algorithm escape from a local minimum.
3. Increase the number of iterations: If you haven't already, try running the algorithm for a larger number of iterations. Sometimes, it takes more iterations for the optimization process to converge to the desired solution.
4. Reevaluate the model: If none of the above steps yield satisfactory results, it might be worth reevaluating the model itself. Are there any assumptions or simplifications that could be affecting the convergence. Consider revising the model structure or exploring alternative modelling approaches.

Hope it helps,

Best Regards,

Raghav Bansal

MathWorks Technical Support